Add execution windows to runtime target

Author: Pigbibi

docs/platform_runtime_inventory.md

@@ -1,6 +1,6 @@
 # Platform Runtime Inventory
 
-_Verified snapshot: 2026-04-18_
+_Verified snapshot: 2026-05-14_
 
 This document records the public runtime wiring inventory across platform repositories and deployment projects. It is meant to answer one question quickly:
 
@@ -45,7 +45,12 @@ For the platform / strategy-domain / configurable-profile matrix, see [`platform
 - **Runtime service account**
   - `ibkr-platform-runtime@interactivebrokersquant.iam.gserviceaccount.com`
 - **Scheduler**
-  - `interactive-brokers-quant-service-scheduler`
+  - `interactive-brokers-live-slot-a-precheck-scheduler`
+  - `interactive-brokers-live-slot-a-probe-scheduler`
+  - `interactive-brokers-live-slot-a-scheduler`
+  - `interactive-brokers-live-slot-b-precheck-scheduler`
+  - `interactive-brokers-live-slot-b-probe-scheduler`
+  - `interactive-brokers-live-slot-b-scheduler`
   - region: `us-central1`
 - **Core runtime selectors**
   - `STRATEGY_PROFILE=<runtime_enabled us_equity profile>`
@@ -69,6 +74,8 @@ For the platform / strategy-domain / configurable-profile matrix, see [`platform
 - **Runtime service account**
   - `schwab-platform-runtime@charlesschwabquant.iam.gserviceaccount.com`
 - **Scheduler**
+  - `charles-schwab-quant-service-precheck-scheduler`
+  - `charles-schwab-quant-service-probe-scheduler`
   - `charles-schwab-quant-service-scheduler`
   - region: `us-central1`
 - **Core runtime selectors**
@@ -97,19 +104,27 @@ For the platform / strategy-domain / configurable-profile matrix, see [`platform
 - **Runtime service account**
   - `longbridge-platform-runtime@longbridgequant.iam.gserviceaccount.com`
 - **Schedulers**
+  - `longbridge-quant-paper-service-precheck-scheduler` in `asia-east2`
+  - `longbridge-quant-paper-service-probe-scheduler` in `asia-east2`
   - `longbridge-quant-paper-service-scheduler` in `asia-east2`
-  - HK: reserved / not yet wired
+  - `longbridge-quant-hk-service-precheck-scheduler` in `asia-east2`
+  - `longbridge-quant-hk-service-probe-scheduler` in `asia-east2`
+  - `longbridge-quant-hk-service-scheduler` in `asia-east2`
+  - `longbridge-quant-sg-service-precheck-scheduler` in `asia-southeast1`
+  - `longbridge-quant-sg-service-probe-scheduler` in `asia-southeast1`
   - `longbridge-quant-sg-service-scheduler` in `asia-southeast1`
 - **Core runtime selectors**
-  - `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` per account service
+- `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` per account service
   - `ACCOUNT_REGION=PAPER|HK|SG`
   - `LONGPORT_SECRET_NAME=<account token secret>`
 - **Runtime secrets**
   - Secret Manager refs for LongPort app key / app secret
   - account token secrets selected by `LONGPORT_SECRET_NAME`
   - runtime Telegram token secret
 - **Runtime notes**
-  - PAPER and SG are live today; HK keeps the same deployment pattern when it is added. Each account identity gets its own Cloud Run service, trigger, and GitHub Environment.
+  - Each account identity gets its own Cloud Run service, trigger, and GitHub Environment. HK can keep the same deployment pattern when it is promoted; rollout state is controlled outside the public docs.
+- Each live account identity now has three Cloud Scheduler triggers: `precheck` after the open window, `probe` at open+30 minutes, and the root execution trigger near the close window.
+- For Cloud Scheduler HTTP targets, keep the OIDC audience on the Cloud Run service root URL; do not append `precheck` or `probe` to the audience.
   - Snapshot-backed profiles require feature snapshot path / manifest envs; direct-runtime profiles do not.
   - App key / secret are account-specific Secret Manager refs; Telegram token is shared inside the LongBridge project.
   - `SERVICE_NAME` should use the full runtime-facing service names above, not older short prefixes.

docs/platform_runtime_inventory.zh-CN.md

@@ -1,6 +1,6 @@
 # 平台运行接线清单
 
-_校验快照日期：2026-04-18_
+_校验快照日期：2026-05-14_
 
 这份文档记录公开仓库里可以保留的 runtime 接线信息，用来快速回答一个问题：
 
@@ -44,7 +44,12 @@ _校验快照日期：2026-04-18_
 - **runtime service account**
   - `ibkr-platform-runtime@interactivebrokersquant.iam.gserviceaccount.com`
 - **Scheduler**
-  - `interactive-brokers-quant-service-scheduler`
+  - `interactive-brokers-live-slot-a-precheck-scheduler`
+  - `interactive-brokers-live-slot-a-probe-scheduler`
+  - `interactive-brokers-live-slot-a-scheduler`
+  - `interactive-brokers-live-slot-b-precheck-scheduler`
+  - `interactive-brokers-live-slot-b-probe-scheduler`
+  - `interactive-brokers-live-slot-b-scheduler`
   - region：`us-central1`
 - **核心运行选择器**
   - `STRATEGY_PROFILE=<runtime_enabled us_equity profile>`
@@ -68,6 +73,8 @@ _校验快照日期：2026-04-18_
 - **runtime service account**
   - `schwab-platform-runtime@charlesschwabquant.iam.gserviceaccount.com`
 - **Scheduler**
+  - `charles-schwab-quant-service-precheck-scheduler`
+  - `charles-schwab-quant-service-probe-scheduler`
   - `charles-schwab-quant-service-scheduler`
   - region：`us-central1`
 - **核心运行选择器**
@@ -95,7 +102,14 @@ _校验快照日期：2026-04-18_
 - **runtime service account**
   - `longbridge-platform-runtime@longbridgequant.iam.gserviceaccount.com`
 - **Scheduler**
+  - `longbridge-quant-paper-service-precheck-scheduler`（`asia-east2`）
+  - `longbridge-quant-paper-service-probe-scheduler`（`asia-east2`）
+  - `longbridge-quant-paper-service-scheduler`（`asia-east2`）
+  - `longbridge-quant-hk-service-precheck-scheduler`（`asia-east2`）
+  - `longbridge-quant-hk-service-probe-scheduler`（`asia-east2`）
   - `longbridge-quant-hk-service-scheduler`（`asia-east2`）
+  - `longbridge-quant-sg-service-precheck-scheduler`（`asia-southeast1`）
+  - `longbridge-quant-sg-service-probe-scheduler`（`asia-southeast1`）
   - `longbridge-quant-sg-service-scheduler`（`asia-southeast1`）
 - **核心运行选择器**
   - 每个区域服务设置 `STRATEGY_PROFILE=<runtime_enabled us_equity profile>`
@@ -106,7 +120,7 @@ _校验快照日期：2026-04-18_
   - 由 `LONGPORT_SECRET_NAME` 选择的区域 token secret
   - runtime Telegram token secret
 - **运行说明**
-  - HK / SG 继续保持两个 Cloud Run 服务、两个 trigger、两个 GitHub Environment。
+  - 每个账户身份都对应自己的 Cloud Run 服务、trigger 和 GitHub Environment；公开文档不记录当前哪几个账户处于实盘。每个账户都有三条 Cloud Scheduler 触发：开盘后 `precheck`，开盘后 `probe`（+30 分钟），临近收盘执行。Scheduler 的 OIDC audience 也要指向 Cloud Run 服务根 URL，不要拼到 `precheck` 或 `probe` 路径。
   - snapshot 驱动策略需要 feature snapshot path / manifest env；直接运行输入策略不需要。
   - App key / secret 使用分区域的 Secret Manager 引用；Telegram token 在 LongBridge 项目内共享。
   - `SERVICE_NAME` 应使用上面的完整运行时服务名，不再使用旧短前缀。

docs/platform_strategy_matrix.md

@@ -29,7 +29,7 @@ For strategy behavior, research status, and archived backtest evidence, see
 |---|---|---|---|---|---|---|
 | IBKR | `QuantStrategyLab/InteractiveBrokersPlatform` | `us_equity` | `RUNTIME_TARGET_JSON` + `ACCOUNT_GROUP` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | Yes - controlled by platform rollout config |
 | Charles Schwab | `QuantStrategyLab/CharlesSchwabPlatform` | `us_equity` | `RUNTIME_TARGET_JSON` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | Yes - controlled by platform rollout config |
-| LongBridge | `QuantStrategyLab/LongBridgePlatform` | `us_equity` | `RUNTIME_TARGET_JSON` + `ACCOUNT_REGION` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` per account service | Cloud Run | Yes - paper and SG today; HK later |
+| LongBridge | `QuantStrategyLab/LongBridgePlatform` | `us_equity` | `RUNTIME_TARGET_JSON` + `ACCOUNT_REGION` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` per account service | Cloud Run | Yes - controlled by per-account rollout config |
 | Binance | `QuantStrategyLab/BinancePlatform` | `crypto` | `RUNTIME_TARGET_JSON` (workflow-local) | `crypto_leader_rotation` | Oracle Cloud + self-hosted runner | No - only this profile is supported today |
 
 ## What this means right now
@@ -42,7 +42,7 @@ Platforms currently in this domain:
 - `CharlesSchwabPlatform`
 - `LongBridgePlatform`
 
-LongBridge account identities are modeled as `paper`, `HK`, and `SG`; today the live services use `paper` and `SG`, and `HK` follows the same contract when it is introduced.
+LongBridge account identities are modeled as `paper`, `HK`, and `SG`; each account service follows the same contract, with deployment-specific rollout control deciding which ones are active.
 
 Important limitation:
 

docs/platform_strategy_matrix.zh-CN.md

@@ -29,7 +29,7 @@ _核对时间：2026-04-18_
 |---|---|---|---|---|---|
 | IBKR | `QuantStrategyLab/InteractiveBrokersPlatform` | `us_equity` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | 可以，按平台 rollout 配置启用 |
 | Charles Schwab | `QuantStrategyLab/CharlesSchwabPlatform` | `us_equity` | `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | 可以，按平台 rollout 配置启用 |
-| LongBridge | `QuantStrategyLab/LongBridgePlatform` | `us_equity` | 每个区域服务配置 `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | 可以，按平台 rollout 配置启用 |
+| LongBridge | `QuantStrategyLab/LongBridgePlatform` | `us_equity` | 每个区域服务配置 `STRATEGY_PROFILE=<runtime_enabled us_equity profile>` | Cloud Run | 可以，按账户服务 rollout 配置启用 |
 | Binance | `QuantStrategyLab/BinancePlatform` | `crypto` | `crypto_leader_rotation` | Oracle Cloud + self-hosted runner | 当前只支持这个 crypto profile |
 
 ## 这张表现在该怎么理解

src/quant_platform_kit/common/runtime_target.py

@@ -15,6 +15,7 @@ class RuntimeTarget:
     account_selector: tuple[str, ...] = ()
     account_scope: str | None = None
     service_name: str | None = None
+    execution_windows: dict[str, Any] | None = None
 
     @property
     def execution_mode(self) -> str:
@@ -67,6 +68,7 @@ def build_runtime_target(
     account_selector: Iterable[str] | str | None = None,
     account_scope: str | None = None,
     service_name: str | None = None,
+    execution_windows: Mapping[str, Any] | None = None,
 ) -> RuntimeTarget:
     return RuntimeTarget(
         platform_id=str(platform_id).strip(),
@@ -76,6 +78,7 @@ def build_runtime_target(
         account_selector=_normalize_account_selector(account_selector),
         account_scope=_normalize_optional_string(account_scope),
         service_name=_normalize_optional_string(service_name),
+        execution_windows=dict(execution_windows) if execution_windows is not None else None,
     )
 
 
@@ -137,6 +140,10 @@ def resolve_runtime_target_from_env(
                 "RUNTIME_TARGET_JSON.execution_mode does not match dry_run_only"
             )
 
+    execution_windows = payload.get("execution_windows")
+    if execution_windows is not None and not isinstance(execution_windows, dict):
+        raise ValueError("RUNTIME_TARGET_JSON.execution_windows must be an object when present")
+
     return build_runtime_target(
         platform_id=resolved_platform_id,
         strategy_profile=resolved_strategy_profile,
@@ -145,4 +152,5 @@ def resolve_runtime_target_from_env(
         account_selector=payload.get("account_selector"),
         account_scope=payload.get("account_scope"),
         service_name=payload.get("service_name"),
+        execution_windows=execution_windows,
     )

tests/test_runtime_target.py

@@ -19,6 +19,10 @@ def test_build_runtime_target_normalizes_selectors_and_mode(self) -> None:
             account_selector=(" U123 ", "", None),
             account_scope=" hk ",
             service_name=" longbridge-quant-hk-service ",
+            execution_windows={
+                "precheck": {"enabled": True, "offset_minutes": 15, "mode": "notify_only"},
+                "execution": {"enabled": True, "offset_minutes": 15, "mode": "paper"},
+            },
         )
 
         self.assertEqual(target.platform_id, "longbridge")
@@ -28,6 +32,8 @@ def test_build_runtime_target_normalizes_selectors_and_mode(self) -> None:
         self.assertEqual(target.account_selector, ("U123",))
         self.assertEqual(target.account_scope, "hk")
         self.assertEqual(target.service_name, "longbridge-quant-hk-service")
+        self.assertEqual(target.execution_windows["precheck"]["offset_minutes"], 15)
+        self.assertEqual(target.execution_windows["execution"]["mode"], "paper")
 
     def test_build_runtime_target_supports_live_mode_without_account_selector(self) -> None:
         target = build_runtime_target(
@@ -50,7 +56,9 @@ def test_resolve_runtime_target_from_env_prefers_structured_json(self) -> None:
                     '{"platform_id":"longbridge","strategy_profile":"global_etf_rotation",'
                     '"dry_run_only":true,"deployment_selector":"HK","account_selector":["HK"],'
                     '"account_scope":"HK","service_name":"longbridge-quant-hk-service",'
-                    '"execution_mode":"paper"}'
+                    '"execution_mode":"paper","execution_windows":{"precheck":{"enabled":true,'
+                    '"offset_minutes":15,"mode":"notify_only"},"execution":{"enabled":true,'
+                    '"offset_minutes":15,"mode":"paper"}}}'
                 )
             },
         )
@@ -63,6 +71,8 @@ def test_resolve_runtime_target_from_env_prefers_structured_json(self) -> None:
         self.assertEqual(target.account_selector, ("HK",))
         self.assertEqual(target.account_scope, "HK")
         self.assertEqual(target.service_name, "longbridge-quant-hk-service")
+        self.assertTrue(target.execution_windows["precheck"]["enabled"])
+        self.assertEqual(target.execution_windows["execution"]["offset_minutes"], 15)
 
     def test_resolve_runtime_target_from_env_rejects_mismatched_execution_mode(self) -> None:
         with self.assertRaisesRegex(ValueError, "execution_mode does not match dry_run_only"):
@@ -101,6 +111,10 @@ def test_build_runtime_context_fields_merges_runtime_target_without_overwriting_
             strategy_profile="soxl_soxx_trend_income",
             dry_run_only=True,
             service_name="longbridge-platform",
+            execution_windows={
+                "precheck": {"enabled": True, "offset_minutes": 15, "mode": "notify_only"},
+                "execution": {"enabled": True, "offset_minutes": 15, "mode": "paper"},
+            },
         )
 
         fields = build_runtime_context_fields(
@@ -115,3 +129,4 @@ def test_build_runtime_context_fields_merges_runtime_target_without_overwriting_
         self.assertEqual(fields["account_scope"], "HK")
         self.assertEqual(fields["runtime_target"]["platform_id"], "longbridge")
         self.assertEqual(fields["runtime_target"]["execution_mode"], "paper")
+        self.assertEqual(fields["runtime_target"]["execution_windows"]["precheck"]["enabled"], True)